Data-science driven autonomous process optimization.

Area of Science:

• Chemical Engineering
• Organic Chemistry
• Process Chemistry

Background:

• Autonomous process optimization offers human intervention-free exploration of reaction parameters.
• Achieving high product yield and selectivity is a key goal in chemical synthesis.
• Traditional methods for selecting categorical parameters, such as phosphine ligands, can introduce bias.

Purpose of the Study:

• To develop and implement a closed-loop system for parallel autonomous process optimization in batch.
• To address the critical aspect of defining meaningful, broad, and unbiased process parameters.
• To establish a systematic strategy for selecting diverse phosphine ligands, overcoming reliance on chemical intuition.

Main Methods:

• Development of a closed-loop system using off-the-shelf components for autonomous experiments.
• Implementation of the system for optimizing a stereoselective Suzuki-Miyaura coupling reaction.
• Leveraging computed molecular feature clustering for systematic selection of phosphine ligands.

Main Results:

• The developed system enables parallel autonomous process optimization.
• Identification of phosphine ligand as a vital categorical parameter influencing reaction outcomes.
• A novel strategy using molecular feature clustering for unbiased ligand selection was established.

Conclusions:

• Systematic, unbiased parameter selection is crucial for successful autonomous optimization.
• Computed molecular feature clustering provides a robust method for categorical parameter selection.
• The optimized Suzuki-Miyaura coupling achieved selective access to the desired product isomer in high yield.